Hosting Services — The Short and Mid-Term Solution Before Transition to the Public Cloud

8 min readJust now

Anyone who follows my posts on social networks knows that I am an advocate of the public cloud and cloud adoption.

According to Synergy Research Group, by 2029, “hyperscale operators will account for over 60% of all capacity, while on-premise will drop to just 20%”.

In this blog post, I will share my personal opinion on where I see the future of IT infrastructure.

Before we begin the conversation, I believe we can all agree that many organizations are still maintaining on-prem data centers and legacy applications, and they are not going away any time in the foreseeable future.

Another thing I hope we can agree on is that IT infrastructure aims to support the business, but it does not produce direct revenue for the organization.

Now, let us talk about how I see the future of IT infrastructure in the short or mid-term, and in the long term.

Short or Mid-term future — Hosting or co-location services

Many organizations are still maintaining on-prem data centers of various sizes. There are many valid reasons to keep maintaining on-prem data centers, to name a few:

Keeping investment in purchased hardware (physical servers, network equipment, storage arrays, etc.)
Multi-year software license agreements to commercial vendors (such as virtualization, operating systems, databases, etc.)
Legacy application constraints (from license bound to a physical CPU to monolith applications that were developed many years ago and simply were not developed to run in cloud environments, least not efficient)
Regulatory constraints requiring to keep data (such as customer data) in a specific country/region.
Large volumes of data are generated and stored for many years, and the cost/time constraints to move it to the public cloud.
Employee knowledge — many IT veterans have already invested years over years learning how to deploy hardware/virtualization, how to maintain network or storage equipment, and let us be honest — they may be afraid of making a change, learning something new such as moving to the public cloud.

Regardless of how organizations see on-prem data centers, they will never have the expertise that large-scale hosting providers have, to name a few:

Ability to deploy high-scale and redundant data centers, in multiple locations (think about primary and DR data centers or active-active data centers)
Invest in physical security (i.e., who has physical access to the data center and specific cage), while providing services to multiple different customers.
Build and maintain sustainable data centers, using the latest energy technologies, while keeping the carbon footprint to the minimum.
Ability to recruit highly skilled personnel (from IT, security, DevOps, DBA, etc.) to support multiple customers of the hosting service data centers.

What are the benefits of hosting services?

Organizations debating about migration to the cloud, and the efforts required to re-architect or adjust their applications and infrastructure to make them efficient to run in the public cloud, could use hosting services.

Here are a few examples for using hosting services:

Keeping legacy hardware that still produces value for the business, such as Mainframe servers (for customers such as banks or government agencies). Organizations would still be able to consume their existing applications, without having to take care of the ongoing maintenance of legacy hardware.
Cloud local alternative on-prem — Several CSPs are offering their hardware racks for local or hosting facilities deployment, such as AWS Outposts, Azure Local, or Oracle Cloud@Customer, allowing customers to consume hardware identical to the hardware that the CSPs offer (managed remotely by the CSPs), while using the same APIs, but locally.
Many organizations are limited by their Internet bandwidth. By using hosting services, organizations can leverage the network bandwidth of their hosting provider, and the ability to have multiple ISPs, allowing them both inbound and outbound network connectivity.
Many organizations would like to begin developing GenAI applications, which requires them an invest in expensive GPU hardware. The latest hardware costs a lot of money, requires dedicated cooling and it gets outdated over time. Instead, hosting service provider can purchase and maintain a large number of GPUs, offering their customers the ability to consume GPU to AI/ML workloads, while paying for the time they used the hardware.
Modern pricing model — Hosting services can offer customers modern pricing plans, such as paying for the actual amount of storage consumed, paying for the time a machine was running (or offering to purchase hardware for several years in advance), Internet bandwidth (or traffic consumed on the network equipment), etc.

Hosting services and the future of modern applications

Just because organizations will migrate from their on-prem to hosting services, does not mean they cannot begin developing modern applications.

Although hosting services do not offer all the capabilities of the public cloud (such as infinite scale, perform actions, and consumption information via API calls, etc.), there are still many things organizations can begin today, to name a few:

Deploy applications inside containers — a hosting provider can deploy and maintain Kubernetes control plane for his customers, allowing them to consume Kubernetes, without having to take care of the burden related to Kubernetes maintenance. A common example of Kubernetes that can be deployed locally at a hosting provider facility, and later on can be used in public cloud environments is OpenShift.
Consume object storage — a hosting provider can deploy and maintain object storage services (such as Min.io), offering his customers to begin consuming storage capabilities that exist in cloud-native environments.
Consume open-source queuing services — customers can deploy message brokers such as ActiveMQ or RabbitMQ, to develop asynchronous applications, and when moving to the public cloud, use the cloud providers managed services alternatives.
Consume message streaming services — customers can begin deploying event-driven architectures using Apache Kafka, to stream a large number of messages, in near real-time, and when moving to the public cloud, use the cloud providers managed services alternatives.
Deploy components using Infrastructure as Code. Some of the common IaC alternatives such as Terraform and Pulumi, already support providers from the on-prem environment (such as Kubernetes), which allows organizations to already begin using modern deployment capabilities.

Note — perhaps the biggest downside of hosting services in the area of modern applications is the lack of function-as-a-service capabilities. Most FaaS are vendor-opinionated, and I have not heard of many customers using FaaS locally.

Transitioning from on-prem to hosting services

The transition between the on-prem and a hosting facility should be straightforward, after all, you are keeping your existing hardware (from Mainframe servers to physical appliances) and simply have the hosting provider take care of the ongoing maintenance.

Hosting services will allow organizations access to managed infrastructure (similar to the Infrastructure as a Service model in the public cloud), and some providers will also offer you managed services (such as storage, WAF, DDoS protection, and perhaps even managed Kubernetes or databases).

Similarly to the public cloud, the concept of shared responsibility is still relevant. The hosting provider is responsible for all the lower layers (from physical security, physical hardware, network, and storage equipment, up until the virtualization layer), and organizations are responsible for whatever happens within their virtual servers (such as who has access, what permissions are granted, what data is being stored, etc.). In case an organization needs to comply with regulations (such as PCI-DSS, FedRAMP, etc.), the organization needs to work with the hosting provider to figure out how to comply with the regulation end-to-end (do not assume that if your hosting provider’s physical layer is compliant, so does your OS and data layers).

Long-term future — The public cloud

I have been a cloud advocate for many years, so my opinion about the public cloud is a little bit biased.

The public cloud brings agility into the traditional IT infrastructure — when designing an architecture, you have multiple ways to achieve similar goals — from traditional architecture based on VMs to modern designs such as microservice architecture built on top of containers, FaaS, or event-driven architecture.

One of the biggest benefits of using the public cloud is the ability to consume managed services (such as managed database services, managed Kubernetes services, managed load-balancers, etc.), where from a customer point of view, you do not need to take care of compute scale (i.e., selecting the underlying compute hardware resources or number of deployed compute instances), or the ongoing maintenance of the underlying layers.

Elasticity and infinite resource scale are huge benefits of the public cloud (at least for the hyper-scale cloud providers), which no data center can compete with. Organizations can design architectures that will dynamically adjust the number of resources according to customers’ load (up or down).

For many IT veterans, moving to the public cloud requires a long learning curve (among others, switching from being a specialist in a certain domain to becoming a generalist in multiple domains). Organizations need to invest resources in employee training and knowledge sharing.

Another important topic that organizations need to focus on while moving to the public cloud is cost — it should be embedded in any architecture or design decision. Engineers and architects need to understand the pricing model for each service they are planning to use and try to select the most efficient service alternative (such as storage tier, compute size, database type, etc.)

Summary

The world of IT infrastructure as we currently know it is constantly changing. Organizations would still like to gain value from their past investment in hardware and legacy applications. In my personal opinion, using legacy applications that still produce value (until they finally reach the decommission phase), simply does not worth the burden of having to maintain on-prem data centers.

Organizations should focus on what brings them value, such as developing new products or providing better services for their customers, and shift the ongoing maintenance to providers who specialize in this field.

For startups who were already born in the cloud, the best alternative is building cloud-native applications in the public cloud.

For traditional organizations, that still maintain legacy hardware and applications, the most suitable alternative for the short or mid-term is to move away from their existing data centers, to one of the hyper-scale or dedicated hosting providers in their local country.

In the long term, organizations should assess all their existing applications and infrastructure, and either decommission old applications, or re-architect / modernize to be deployed in the public cloud, using managed services, modern architectures (microservices, containers, FaaS, event-driven architecture, etc.), using modern deployment methods (i.e., Infrastructure as Code).

The future will probably be a mix of hyper-clouds (hosting facilities combined with one or more public clouds), and single or multiple public cloud providers.

About the author

Eyal Estrin is a cloud and information security architect, an AWS Community Builder, and the author of the books Cloud Security Handbook and Security for Cloud Native Applications, with more than 20 years in the IT industry.

You can connect with him on social media (https://linktr.ee/eyalestrin).

Opinions are his own and not the views of his employer.